Class summary: Friday, 3/14

[This material will not be on exams.]
I discussed a variation of a famous paradox, known as the "St. Petersbug paradox" (or "St. Petersburg game"). It involves playing a game with even odds (say, tossing a coin), in which you win or lose $1 (net) for each bet of $1, depending whether you win or lose the game. You are free to set the amount to bet on the game, and to stop playing at any time. With most naive strategies, like betting $1 on each game, your expected net win after a series of games is $0. However, it turns out that there is a strategy that guarantees you a $1 net win eventually, no matter how the individual games turn out; hence your expected net win under this strategy is $1. Namely, you start betting $1 on the first game, and then keep doubling your bet until you win a game, at which point you stop. It is not hard to see that (a) after the win occurs (and you therefore stop) you are ahead by exactly $1, and (b) that the probability that you win a game at some point is 1. Thus, the strategy guarantees you a win with 100 percent probability.

The usual formulation of the St Petersburg paradox is somewhat different, but based on the same idea: It involves a casino game with even odds, and you keep playing this game until you win. At that point, the casino pays you $2n, where n is the number of games you have played. One can show that the expected win is infinite. Hence, you matter how much the casino charges you to play this game, you will come out ahead.

Of course, there are a few catches, that prevent one from exploiting this game in the real world. In particular, the amount needed to bet under the "double-if-you-lose" strategy becomes astronomical after only a few dozen games, so in reality you cannot play arbitrarily many games. However, if the number of games is restricted (say, at most 20 games so that the bet does not exceed $220), then the paradox disappears: the expected net win under such a restriction is 0.

Here are two websites dealing with this paradox:


Back to the Math 361 X1 Homepage